Towards an objective approach for a regional - continental scale geomorphic river classification

River classification schemes are now widely used by a range of government agencies, managers and researchers to help reach an understanding of river form and process among the geomorphic complexity found in river channels, as a basis for understanding ecosystem patch dynamics and connections, and as means of organising and prioritising research and management activities. With some exceptions, existing schemas are generally qualitative, relying largely on expert judgement to delineate “homogeneous geomorphic reaches” for a specific river, from field, GIS, and/or remotely sensed data. In this paper, we demonstrate that continuous data for a number of morphological metrics can be derived relatively cheaply from available remotely sensed and GIS data, as a basis for a regional to continental geomorphic river classification. This approach is intended as a “proof of concept” for further development into a classification approach that can be applied across Northern Australia. Others, have used many of the metrics used here in one form or another, however, they are rarely derived as continuous data throughout the drainage network. A test of the derived metrics against an existing classification within the Mitchell River in the Gulf of Carpentaria suggest that most existing reach classes can be discriminated within the derived metrics. Introduction The geomorphic and applied ecology literature is replete with examples of different river classification schemes, all of which have their own merits and limitations (e.g. Rosgen, 1994, 1996; Miller and Ritter, 1996; Kondolf, 1995). Most of these classifications are inherently subjective and highly influenced by the experience of the classifier, they are also influenced by the resolution of the data available to them and the discontinuous, and often arbitrary, spatial distribution of the field data. The reliance of most classification schemes on field data also limits the practical application of such approaches for regional or continental scale analysis, particularly to remote and inaccessible areas. At present a universally accepted geomorphic river classification scheme does not exist that can be applied at the continental scale, or indeed any scale (Naiman, 1992; Kondolf, 1995, Parsons et.al. 2004). Numerous attempts have been made at developing such schemes, with the Rosgen (1994) and the River Styles approach of Brierley and Fryirs (2005) being recent examples in a line extending back to Davis (1899). Naiman et.al. (1992) state that despite the lack of a universally accepted classification framework (and it should be noted that 15 years on, the situation remains the same) a consensus has been reached regarding the fundamental attributes of an enduring classification scheme. They state; “it should have the ability to encompass broad spatial and temporal scales, to integrate structural and functional characteristics under various in-stream disturbance regimes, to convey information about underlying mechanisms controlling in-stream features, and to accomplish this at low cost and at a high level of understanding among resource managers.” (Naiman et al., 1992, p117). To this should also be added that the classification scheme should be as objective as possible and should be repeatable by different operators applying the classification system independently (Kondolf, 1995). A common property of many river classification schemes is that they are dependent on high resolution field data, which are often collected at non randomly selected sites, frequently dictated by accessibility (see Parsons, et al., 2004). Accessibility issues and spatial scale in the remote areas of Northern Australia severely limit the ability to use field sampling to detect major transitions. Several researchers have utilised multivariate statistical techniques as a means of removing much of the subjectivity from classification (e.g. Mosley, 1987, Newson et al., 1998, Thorp et al., 2006) with varying degrees of success. Mosley (1987) was unable to distinguish meaningful classes from 190 river reaches in New Zealand, a result that may have been a function of the discontinuous nature of the data being used. The multivariate approaches reported by Newson et al., 1998 and Thorp et al. 2006, show more promise, providing the appropriate scale input data is available.